Component Health Monitoring

The present disclosure relates generally to health monitoring for aircraft components and in particular, to performing health monitoring of aircraft components using neural networks.

Health monitoring is performed for aircraft to increase safety, increase operational efficiency, and reduce costs. This monitoring can be performed for aircraft components. This monitoring can be for aircraft components such as generators, engines, avionics systems, and other components. This monitoring can be performed continuously during operation of an aircraft.

With this type of monitoring, potential inconsistencies or out of tolerance operations of aircraft components can be detected. By detecting these potential inconsistencies or out of tolerance operations, maintenance can be performed prior to an aircraft being out of a desired operating tolerance. Further, this type of monitoring can also be used by airlines to make decisions about fleet management, component replacement, operational planning, and other actions regarding aircraft.

An embodiment of the present disclosure provides a generator health monitoring system for a generator. The generator bearing health monitoring system comprises a computer system, neural network in the computer system, and health analyzer in the computer system. The neural network comprises layers, an input layer in the layers in the neural network, and a last layer before an output layer in the layers in the neural network. The input layer is configured to receive sensor data for a set of variables for generator bearings in the generator. The last layer is configured to receive condition indicator data for a set of condition indicators. The neural network is trained to predict a health status of the generator bearings using the sensor data for the set of variables and the condition indicator data for the set of condition indicators. The health analyzer is configured to identify the sensor data for the set of variables from a sensor system monitoring the generator including the generator bearings; identify the condition indicator data for the set of condition indicators for the generator bearings; input the sensor data into the input layer; input the condition indicator data into the last layer; and receive the prediction of the health status of the generator bearings from the output layer for the neural network in response to inputting the sensor data and the condition indicator data.

An embodiment of the present disclosure provides a health monitoring system comprising a computer system, a neural network in the computer system, and a health analyzer in the computer system. The neural network comprises layers, an input layer in the layers in the neural network, and a subsequent layer in the layers in the neural network. The input layer is configured to receive sensor data for a set of variables for a component. The subsequent layer is configured to receive condition indicator data for a set of condition indicators. The neural network is trained to predict a health status of the component using the sensor data for the set of variables and the condition indicator data for the set of condition indicators. The health analyzer is configured to: input the sensor data into the input layer; input the condition indicator data into the subsequent layer; and receive the health status predicted for the component from the output layer in response to inputting the sensor data and the condition indicator data.

Yet another embodiment of the present disclosure provides a method for monitoring a health of generator bearings. Sensor data is identified for a set of variables from a sensor system monitoring a generator including the generator bearings. Condition indicator data for a set of condition indicators for the generator bearings is identified. The sensor data is input into an input layer in layers in a neural network. The condition indicator data is input into a last layer before an output layer in the layers in the neural network. The neural network is trained to predict a health status of the generator bearings using the sensor data for the set of variables and the condition indicator data for the set of condition indicators. A prediction of the health status for the generator bearings is received from the output layer in response to inputting the sensor data and the condition indicator data.

Still another embodiment of the present disclosure provides a method for monitoring a health of a component. Sensor data for a set of variables for the component is input into an input layer in layers in a neural network. Condition indicator data for a set of condition indicators is input into a subsequent layer before an output layer in the layers in the neural network. The neural network is trained to predict a health status of the component using the sensor data for the set of variables and the condition indicator data for the set of condition indicators. The health status predicted for the component is received from the output layer in response to inputting the sensor data and the condition indicator data.

The features and functions can be achieved independently in various embodiments of the present disclosure or may be combined in yet other embodiments in which further details can be seen with reference to the following description and drawings.

The illustrative embodiments recognize and take into account one or more different considerations as described herein. For example, currently used condition-based maintenance methods for aircraft components are based on statistics and trending and leverage domain experts and failure instances to establish thresholds. The experts select the relevant data to implement condition indicators. Further, the current techniques do not use a neural network. In addition, current techniques do not incorporate expert knowledge with the model. This expert knowledge can include a selection of condition indicators that may be helpful in predicting the health of an aircraft component.

A machine learning model such as a neural network can be created that takes into account both the sensor data for the aircraft component and expert knowledge. In these illustrative examples, the expert knowledge takes the form of condition indicators that have been identified as useful in determining the health of aircraft components but are not definitive for use in determining the health of the aircraft components.

In one illustrative example, analog data is sampled with high frequency in the form of multivariate time series in an aircraft component such as generator bearings in a generator. This data is sensor data for a set of variables relating to the generator bearings. The sensor data is used to determine health status using a neural network. Further, physics knowledge about the generator bearings is also input into the neural network. This physics knowledge takes the form of condition indicators data for a set of condition indicators.

In this example, the neural network uses the sensor data and the condition indicator data as inputs and outputs as a health classification for generator bearings in the generator using this data. In one illustrative example, this health classification can be a determination of whether the generator bearings have grease or a lack of grease.

The illustrative examples provide a method, apparatus, system, and computer program product for performing health monitoring. In the illustrative examples, a physics informed deep learning approach can be used to train machine learning models such as neural networks to predict the health of aircraft components. These aircraft components can include generators, engines, landing gear, computer systems, structures, and other aircraft components. In one illustrative example, a physics informed deep learning approach can be used to leverage sensor data and embed physics knowledge of generator bearings into training machine learning model such as a neural network.

For example, with respect to a component such as generator bearings, a generator bearing health monitoring system comprises a computer system, neural network in the computer system, and health analyzer in the computer system. The neural network comprises layer, an input layer in the layers, and a last layer before an output layer in the layers. The input layer is configured to receive sensor data for a set of variables for generator bearings in the generator. The last layer is configured to receive condition indicator data for a set of condition indicators. The neural network is trained to predict a health status of the generator bearings using the sensor data for the set of variables and the condition indicator data for the set of condition indicators.

The health analyzer is configured to receive the sensor data for the set of variables from a sensor system monitoring the generator including the generator bearings; receive the condition indicator data for the set of condition indicators for the generator bearings; input the sensor data into the input layer; input the condition indicator data into the last layer; and receive the prediction of the health status of the generator bearings from the output layer for the neural network.

With reference now to the figures, and in particular, with reference to, an illustration of an aircraft is depicted in accordance with an illustrative embodiment. In this illustrative example, aircrafthas wingand wingattached to body. Aircraftincludes engineattached to wingand engineattached to wing.

Bodyhas tail section. Horizontal stabilizer, horizontal stabilizer, and vertical stabilizerare attached to tail sectionof body.

Aircraftis an example of an aircraft in which health monitoring systemcan be implemented in accordance with an illustrative embodiment. In this illustrative example, health monitoring systemcan monitor the health of various aircraft components in aircraft. This monitoring can be performed continuously, periodically, or in response to events.

For example, health monitoring systemcan operate to continuously monitor the health of generator. Generatorcan be an auxiliary power unit for aircraft. In this example, health monitoring systemcan monitor the health of generator bearings in generator.

With reference now to, an illustration of a block diagram of a health monitoring environment is depicted in accordance with an illustrative embodiment. In this illustrative example, health monitoring environmentincludes components that can be implemented in health monitoring systemin aircraftin.

As depicted, health monitoring systemcan operate to monitor platform. Platformcan take a number of different forms. For example, platformcan be selected from a group comprising a mobile platform, a stationary platform, a land-based structure, an aquatic-based structure, a space-based structure, an aircraft, a commercial aircraft, a rotorcraft, a tilt-rotor aircraft, a tilt wing aircraft, a vertical takeoff and landing aircraft, an electrical vertical takeoff and landing vehicle, a personal air vehicle, a surface ship, a tank, a personnel carrier, a train, a spacecraft, a space station, a satellite, a submarine, an automobile, a power plant, a bridge, a dam, a house, a manufacturing facility, a building, or other types of platforms.

As used herein, “a number of” when used with reference to items, means one or more items. For example, “a number of different forms” is one or more different forms.

In this example, health monitoring systemcan operate to monitor the health of componentin platform. As used herein, a “set of” when used with reference items means one or more items. In this illustrative example, componentcan take a number of different forms. For example, componentcan be selected from a group comprising a generator, generator bearings, a pump, a cooling system, a heat exchanger, an auxiliary power unit, a landing gear system, a wing, an in-flight entertainment system, a computer, or other types of components.

In this illustrative example, health monitoring systemcomprises a number of different components. As depicted, health monitoring systemcomprises computer system, neural network, and health analyzer.

Neural networkis a model comprising nodes organized into layers. These layers in neural networkoperate to perform transformations on input data. These inputs can travel as signals through layers. In this example, layersare fully connected layers. This means that each node in a given layer is connected to every node in a subsequent layer.

In this example, layersinclude input layer, hidden layers, and output layer. Hidden layersare layerslocated between input layerand output layer.

In this example, input layeris connected to first layerin hidden layers. In other words, first layeris configured to receive sensor datafor the set of variablesfor componentthrough input layer. The set of variablescan be selected from at least one of a voltage, a voltage phase, an acceleration or a vibration frequency, a current, a temperature, an acoustic wave, or other suitable variable.

In this example, neural networkis referred to as a deep learning neural network because of the presence of two or more of hidden layers.

In this example, input layerin layersin neural networkis to receive sensor datafor a set of variablesfor component. Sensor systemis configured to generate sensor data. In this depicted example, sensor datais generated from monitoring component. Thus, sensor datacan be used for monitoring or analysis of component. Sensor systemis comprised of a number of sensors. These sensors can include at least one of a temperature sensor, a light sensor, an accelerometer, a pressure sensor, a magnetic sensor, a voltage sensor, a current sensor, or other suitable types of sensors.

Further, the phrase “at least one of,” when used with a list of items, means different combinations of one or more of the listed items can be used, and only one of each item in the list may be needed. In other words, “at least one of” means any combination of items and number of items may be used from the list, but not all of the items in the list are required. The item can be a particular object, a thing, or a category.

For example, without limitation, “at least one of item A, item B, or item C” may include item A, item A and item B, or item B. This example also may include item A, item B, and item C or item B and item C. Of course, any combination of these items can be present. In some illustrative examples, “at least one of” can be, for example, without limitation, two of item A; one of item B; and ten of item C; four of item B and seven of item C; or other suitable combinations.

Sensor datacan take a number of different forms. For example, sensor datacan be selected from at least one of analog time series data or digital time series data. This data can be received in channels in which each channel represents data for a particular variable as generated by a particular sensor in sensors.

In this example, subsequent layeris a layer within hidden layers. In one example, subsequent layercan be last layerin hidden layersbefore output layerin layersin neural network. In this example, subsequent layeris configured to receive condition indicator datafor a set of condition indicators.

In these illustrative examples, physics-based features that are identified as being important with respect to health statusfor componentcan be represented using a set of condition indicators. In other words, the set of condition indicatorscan be selected as physics features identified as being relevant to the health of component. In these examples, condition indicatorscan be derived from sensor datafor the set of variablesfor component. In this illustrative example, a condition indicator is a value derived from sensor dataand can represent physical or operational parameters of component. In this illustrative example, the set of condition indicatorscan be at least one condition indicator thought to be relevant to generator. For example, the set of condition indicatorscan be selected from at least one of a bearing based energy, a ball energy, an inner race energy, an outer race energy, a generator frequency of vibrations for the generator, hydraulic pump frequency, a hydraulic pump piston pass frequency, side lube pump frequency, or other type of condition indicator.

In this illustrative example, condition indicator datafor the set of condition indicatorsis generated using sensor data. The sensor data used to determine condition indicator datacan include data for other variables in variablesin addition to or in place of those variables in the set of variables. For example, sensor datacan include the speed of a shaft for a variable in the form of shaft speed for componentwhen componenttakes the form of generator bearingsin generator. The condition indicator can be a speed threshold for a shaft indicating when the shaft speed is greater than desired based on the specification. For example, if the speed of the shaft is over 100 revolutions per minute (RPM) the condition indicator can be set to 1. If the speed of the shaft is equal to or less than 100 revolutions per minute, the condition indicator can be set to 0.

In other examples, the condition indicator can be determined through more complex calculations. Further, the condition indicator can be determined using multiple variables in variables.

With the use of condition indicator datathat is input into subsequent layerafter first layer, neural networkis physics informed neural network.

In this example, neural networkhas been trained to predict health statusof componentusing sensor datafor the set of variablesand the condition indicator datafor the set of condition indicators. In this example, this prediction is output from output layerin response to receiving sensor dataand condition indicator data.

Health statuspredicted by neural networkcan take a number of different forms. In one illustrative example, health statuscan be selected from a group comprising normal, caution, and warning. This status is for component. In one example, componentcomprises generator bearingslocated in generator.

Health analyzercan be implemented in software, hardware, firmware, or a combination thereof. When software is used, the operations performed by health analyzercan be implemented in program instructions configured to run on hardware, such as a processor unit. When firmware is used, the operations performed by health analyzercan be implemented in program instructions and data and stored in persistent memory to run on a processor unit. When hardware is employed, the hardware can include circuits that operate to perform the operations in health analyzer.

In the illustrative examples, the hardware can take a form selected from at least one of a circuit system, an integrated circuit, an application-specific integrated circuit (ASIC), a programmable logic device, or some other suitable type of hardware configured to perform a number of operations. With a programmable logic device, the device can be configured to perform the number of operations. The device can be reconfigured at a later time or can be permanently configured to perform the number of operations. Programmable logic devices include, for example, a programmable logic array, a programmable array logic, a field-programmable logic array, a field-programmable gate array, and other suitable hardware devices. Additionally, the processes can be implemented in organic components integrated with inorganic components and can be comprised entirely of organic components excluding a human being. For example, the processes can be implemented as circuits in organic semiconductors.

Computer systemis a physical hardware system and includes one or more data processing systems. When more than one data processing system is present in computer system, those data processing systems are in communication with each other using a communications medium. The communications medium can be a network. The data processing systems can be selected from at least one of a computer, a server computer, a tablet computer, or some other suitable data processing system.

As depicted, computer systemincludes a number of processor unitsthat are capable of executing program instructionsimplementing processes in the illustrative examples. In other words, program instructionsare computer-readable program instructions.

As used herein, a processor unit in the number of processor unitsis a hardware device and is comprised of hardware circuits such as those on an integrated circuit that respond to and process instructions and program code that operate a computer.

When the number of processor unitsexecutes program instructionsfor a process, the number of processor unitscan be one or more processor units that are in the same computer or in different computers. In other words, the process can be distributed between processor unitson the same or different computers in computer system.

Further, the number of processor unitscan be of the same type or different types of processor units. For example, the number of processor unitscan be selected from at least one of a single core processor, a dual-core processor, a multi-processor core, a general-purpose central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), or some other type of processor unit.

In this illustrative example, health analyzeridentifies sensor datafor the set of variablesfrom sensor systemmonitoring component. Health analyzeridentifies condition indicator datafor the set of condition indicatorsfor component.

Health analyzerinputs sensor datainto input layer. Health analyzeralso inputs condition indicator datainto subsequent layer. In this example, health analyzerreceives prediction of health statusof componentfrom output layerfor neural networkin response to inputting sensor dataand condition indicator data.

In this example, health analyzercan perform a set of actionsusing health statuspredicted for component. The set of actionscan be selected from at least one of logging health status, generating a warning, scheduling maintenance for component; or halting operation of component.

In one illustrative example, health analyzercan train neural network. For example, health analyzercan identify sample sensor datafor the set of variablesand sample condition indicator datafor the set of condition indicators. With this example, health analyzercan generate training datasetusing sample sensor datafor the set of variablesand sample condition indicator datafor the set of condition indicators. Health analyzercan also include labelsin training dataset. Labelscan be generated by health analyzerfor sample sensor dataand sample condition indicator data. Labelscan be determined from knowing the health of componentat the time sample sensor datawas collected. These labels can also be used with sample condition indicator datadetermined from sample sensor data. Health analyzercan then train neural networkusing training dataset.